Abstract: The object samples of classical SVM (support vector machine ) is large and the operation speed of SVM is slow,which is lead to evaluate the mine geological environment by adopting the classical SVM algorithm with high efficient.The classical SVM algorithm is improved,a new mine geological environment evaluation model based on improved SVM algorithm is proposed.Based on the bit compression principle,firstly,the sample data is conducted bit compression;then,the classifier is trained by using the weighted SVM to improve the convergence rate.Taking the actual measured data of a mining area of Jiangxi province,the comparison and analysis between BP neural network algorithm,classical SVM algorithm and improved SVM algorithm are conducted,the results show that:①the error and convergence speed of the evaluation model established by improved SVM algorithm are superior to the error and convergence speed of the evaluation model established by BP neural network algorithm,the output errors of the evaluation models established by classical SVM algorithm and improved SVM algorithm are similar to each other,but the convergence speed of the evaluation model established by the improved SVM algorithm is higher than that of the evaluation model established by classical SVM algorithm;②with the increasing of bit compression digits,the training sample reduced rate of improved SVM algorithm is increased gradually,the output errors of the evaluation model established by the improved SVM algorithm can be keep unchanged under the conditions of reducing the samples,shortening the training time and improving the convergence rate.The study results of the paper can further indicated that the samples can be decreased,besides that，the training rate is also improved under the conditions of establishing the mine geological environment evaluation model by using the improved SVM algorithm proposed in this paper.

Key words: SVM,Mine geological environment, Bit compression, Weighted SVM, BP neural network